Data‐driven evidence for three distinct patterns of amyloid‐β accumulation. (1st February 2022)

• Record Type: Journal Article
• Authors: Collij, Lyduine E.; Salvadó, Gemma; Wottschel, Viktor; Schoenmakers, Pierre; Mutti, Martina; Aksman, Leon M; Wink, Alle Meije; van der Flier, Wiesje M; Scheltens, Philip; Visser, Pieter Jelle; Van Berckel, Bart N.M.; Barkhof, Frederik; Haller, Sven; Gispert, Juan Domingo; Alves, Isadora Lopes
• Is Part Of: Alzheimer's & dementia. Volume 17(2021)Supplement 4
• Page Start: n/a
• Page End: n/a
• Publication Date: 2022-02-01
• Subjects: Alzheimer's disease -- Periodicals; Alzheimer Disease -- Periodicals; Dementia -- Periodicals; Démence; Maladie d'Alzheimer; Périodique électronique (Descripteur de forme); Ressource Internet (Descripteur de forme); 616.83
• DOI: 10.1002/alz.055417
• Languages: English
• ISSNs: 1552-5260

Abstract

Abstract: Background: Amyloid burden staging models assume a single spatial‐temporal progression of amyloid accumulation across subjects, while disease heterogeneity suggests there might be subtypes with distinct trajectories. Here, we apply a data‐driven model which simultaneously resolves subtypes and stages (SuStaIn) 1 to amyloid PET data to assess evidence for Aβ accumulation subtypes. Method: Amyloid PET data of 3010 subjects () were pooled from ALFA+, EMIF‐AD, ABIDE, OASIS, and ADNI. SUVr was extracted for 17 regions. Next, GMM was applied to regional SUVr of cognitively unimpaired subjects per cohort, per tracer, to derive z‐scores. Then, we applied SuStaIn to baseline data to determine the optimal number of subtypes and the most probable subtype/stage per scan. Subtypes and stages were related to the proportion of APOE ‐ε4 carriers and diagnostic groups, mean age, and Centiloid. The final model was also applied to longitudinal scans (>4 years after baseline) from ADNI and OASIS ( N =519) for validation, where we assessed subtype stability and stage progression. Result: SuStaIn identified three subtypes, referred to as Frontal, Parietal and Occipital based on the first regions to show abnormality (). This 3‐subtype model was preferred to a traditional one‐trajectory model () based on the Cross‐Validation Information Criteria. Most subjects displayed no amyloid (stage 0: N =1810, 60.1%) or abnormal amyloid levels across all regions (stage 17: N =282, 9.4%), challenging subtyping. Of the N =835 (27.7%) with strong subtype assignment (>50% probability), the majority was assigned to Frontal (55.0%), followed by Parietal (24.1%), and Occipital subtypes (21.0%). Statistically significant differences included lower age for the Parietal subtype (Frontal:71.9±8.3; Parietal:69.3±9.6; Occipital:72.1±8.1), and distinct proportions of APOE ‐ε4 carriership (Frontal:61.8%, Parietal:57.1%, Occipital:49.4%) and subjects with dementia across subtypes (Frontal:19.7%, Parietal:19.1%, Occipital:31.0%). Older age, APOE ‐ε4 carriership, and Centiloid burden were positively associated with stage independent of subtype. At follow‐up, most subjects (77.9%) maintained baseline subtype assignment, and 65.3% progressed to a later stage within their subtype. Conclusion: While a one‐trajectory model recovers the established pattern of amyloid accumulation, SuStaIn identified three subtypes with distinct associations to AD risk factors. These initial results suggest subtype classification may have clinical relevance and/or improve risk assessment compared to previous staging models.